After swapping an AI API base URL, re-pull the model IDs first

#ai #api #debugging #devtools

A base_url swap can fail in a boring way: the request is routed to a gateway, but the model name you kept from the old provider is not actually available on the new path.

Before debugging the whole app, make the first receipt smaller.

Model ID receipt checklist

Call GET /v1/models against the new base URL.
Copy the exact model ID from that response.
Send one chat completions request with no RAG, tools, streaming, or fallback.
Check the request log for selected model, route, status, tokens, latency, and cost movement.
Only then re-enable retrieval, agents, retries, and production traffic.

This catches the silent mismatch where a RAG pipeline, agent, or SDK looks broken even though the gateway never loaded the model you thought it was using.

TackleKey's starter path separates registration, key creation, the first successful request, and request-log evidence so a developer can verify the model path before scaling traffic.

Starter path:
https://tacklekey.com/start?utm_source=devto&utm_medium=article&utm_campaign=model_id_receipt_after_base_url_swap&utm_content=model-id-receipt-after-base-url-swap-global-api-20260723-v1